Signal processing device, signal processing method, and program

Abstract

A signal processing device (1) comprises: a plurality of observation systems that use observation data of an aircraft (100) to calculate a smoothed value of the position of the aircraft (100) and an error covariance matrix of the smoothed value; and an integration system that uses a weight, which has been set so as to reduce the error covariance matrix, to integrate a plurality of the smoothed values, which have been output from the plurality of observation systems, and estimate the position of the aircraft (100).

Description

Signal processing device, signal processing method, and program

[0001] This disclosure relates to a signal processing device, a signal processing method, and a program.

[0002] As a signal processing device related to this disclosure, for example, a signal processing device is known that estimates the position of a target from the time from when radio waves are emitted toward the target until the reflected waves are received, and from the direction of arrival of the reflected waves. Examples of the configuration of such a signal processing device (hereinafter also referred to as "conventional device") are shown in Figures 16 to 18.

[0003] In Figure 16, the multiple sensors are radars 2 that observe the position of the target. Here, the sensors are radars, and there are N of them, where N is an integer greater than or equal to 2. The following describes the case where the target is an aircraft 100. The radar tracking filter 3 tracks the position of the aircraft 100 based on the observation values ​​transmitted by the radars 2 and determines the smooth position of the aircraft 100. The data integration unit 10 integrates the data output from the N radar tracking filters 3.

[0004] As shown in Figure 17, the radar tracking filter 3 includes a coordinate transformation unit 18, a Kalman gain calculation unit 19, a smoothing value calculation unit 20, and a prediction value calculation unit 21. The coordinate transformation unit 18 transforms from a polar coordinate system to a Cartesian coordinate system. The Kalman gain calculation unit 19 calculates the Kalman gain used when calculating the smoothing value to determine the smooth position of the target in the tracking process. The smoothing value calculation unit 20 calculates the smoothing value. The prediction value calculation unit 21 predicts the target position in the next sampling.

[0005] Furthermore, as shown in Figure 18, the data integration unit 10 is configured to include a smoothed value memory circuit 22, a covariance memory circuit 23, and an estimated value calculation unit 24. The smoothed value memory circuit 22 stores the smoothed values ​​transmitted from the radar tracking filter 3. The covariance memory circuit 23 stores the error covariance matrix of the smoothed values ​​transmitted from the radar tracking filter 3. The estimated value calculation unit 24 integrates the smoothed values ​​based on the smoothed values ​​transmitted from the smoothed value memory circuit 22 and the error covariance matrix of the smoothed values ​​transmitted from the covariance memory circuit 23, and calculates an estimated value indicating the estimated position of the target.

[0006] Incidentally, regarding the configuration of the conventional device, a configuration in which observation data of the aircraft 100 is observed by a plurality of radars 2 is described in, for example, Patent Document 1. Further, regarding the configuration for integrating the smoothed values from the radar tracking filter 3, an algorithm for calculating the maximum likelihood estimate of the covariance matrix under the situation where the covariance matrix is unknown is described in, for example, Non-Patent Document 1.

[0007] Next, an operation example of the conventional device will be described. Hereinafter, the subscript indicating each of the N radars 2 is denoted as "-#n". n is an arbitrary integer from 1 to N (1 ≤ n ≤ N).

[0008] The radar 2-#n observes the azimuth angle θ ａｚ，ｎ and the distance R ｄ，ｎ of the aircraft 100 which is the target. The radar 2-#n transmits the observation data z ａｚ，ｎ，ｌ and the distance R ｄ，ｎ，ｌ including the azimuth angle θ ｎ，ｌ = (θ ａｚ，ｎ，ｌ , R ｄ，ｎ，ｌ ) at the current time lT with a sampling interval of T to the radar tracking filter 3. After observing the observation data for a predetermined number of samplings L, the radar tracking filter 3 outputs an estimated value x ｎ，ｌ tilde of the position of the aircraft 100. For example, the observation data z ｎ，ｌ at the current time lT transmitted from the radar 2-#n to the radar tracking filter 3 is transmitted to the coordinate conversion unit 18 and the Kalman gain calculation unit 19. The coordinate conversion unit 18 obtains the coordinate-converted observation data z' ｎ，ｌ according to the following formulas (1) to (3). In the following formula (1), (x ｎ，ｌ , y ｎ，ｌ ) are the coordinates of the observed position of the aircraft 100 after coordinate conversion. Further, in the following formulas (2) and (3), the coordinates (x ｒｄ，ｎ , y ｒｄ，ｎ ) represent the position coordinates of the radar 2-#n.

[0009] The observed data z'n,l is transmitted from the coordinate transformation unit 18 to the Kalman gain calculation unit 19 and the smoothing value calculation unit 20. Also, at the sampling time (l-1)T, the predicted value x calculated by the predicted value calculation unit 21 is obtained. ｎ，ｌ｜ｌ－１ This is also transmitted from the prediction value calculation unit 21 to the Kalman gain calculation unit 19. Predicted value x ｎ，ｌ｜ｌ－１ This is constructed as a two-dimensional vector consisting of the predicted value of the x component and the predicted value of the y component. The Kalman gain calculation unit 19 calculates the Kalman gain K according to the following equations (4) to (7). ｎ，ｌ Calculate the following equation (4): P ｌ｜ｌ－１ σ is the prediction error covariance matrix calculated by the prediction value calculation unit 21 at sampling time (l-1)T, and H is the observation matrix. In the following equation (6), σ ＡＺ ２ σ is the error covariance of the azimuth angle observations, Ｒ ２ This is the error covariance of the distance observations.

[0010] The Kalman gain K calculated by the Kalman gain calculation unit 19 ｎ，ｌ This is transmitted to the smoothing value calculation unit 20. The smoothing value calculation unit 20 calculates the smoothing value x according to the following formula (8). ｎ，ｌ｜ｌ Calculate the following. In equation (8) below, z' ｎ，ｌ is the coordinate-transformed observation data represented by equation (1) above, and H is the observation matrix represented by equation (5) above.

[0011] Furthermore, the smoothing value calculation unit 20 calculates the Kalman gain K ｎ，ｌ , Prediction error covariance matrix P ｌ｜ｌ－１ And using the observation matrix H of equation (5) above, the smoothing error covariance matrix P is calculated according to equation (9) below. ｎ，ｌ｜ｌ The smooth value x calculated by the smooth value calculation unit 20 is calculated. ｎ，ｌ｜ｌ and the smoothing error covariance matrix P ｎ，ｌ｜ｌ This is transmitted to the predicted value calculation unit 21.

[0012] The prediction value calculation unit 21 calculates the predicted value x for the next sampling at sampling time lT according to the following formulas (10) and (11). ｎ，ｌ＋１｜ｌ The calculated predicted value x is calculated. ｎ，ｌ＋１｜ｌ This is transmitted to the smoothing value calculation unit 20.

[0013] Furthermore, the prediction value calculation unit 21 uses the transition matrix Φ of equation (11) above to calculate the prediction error covariance matrix P at sampling time lT according to equations (12) and (13) below. ｎ，ｌ＋１｜ｌ The prediction error covariance matrix P is calculated. ｎ，ｌ＋１｜ｌ This is transmitted from the prediction value calculation unit 21 to the Kalman gain calculation unit 19. Note that in the following equation (13), σ ｘ ２ This is the drive noise power in the x-axis direction, and σ ｙ ２ This represents the drive noise power in the y-axis direction.

[0014] The radar tracking filter 3 receives the smoothed value x output from the smoothed value calculation unit 20 at the sampling time LT. ｎ，Ｌ｜Ｌ The estimated value of radar 2-#n x ｎ It outputs as a tilde. Furthermore, the radar tracking filter 3 uses the smoothing error covariance matrix P of radar 2-#n. ｎ，Ｌ｜Ｌ The error covariance matrix P of radar 2-#n ｎ Output as a tilde. Estimated value x ｎ The tilde and its error covariance matrix P ｎ The tilde signal is transmitted from the radar tracking filter 3 to the data integration unit 10.

[0015] The data integration unit 10 calculates the estimated value x １ Tilde ~ x Ｎ The tilde is integrated to calculate the estimated value x tilde. Non-patent document 1 states that the estimated value x tilde is calculated according to the following equations (14) to (16). In the following equation (14), ω １ , , ω Ｎ The respective estimated values ​​x １ Tilde, ..., x ＮThe weight multiplied by the tilde, (0 ≤ ω ｎ The sum is "1" in the range ≤ 1, n = 1, ..., N). １ , , ω Ｎ This is pre-set and transmitted to the data integration unit 10.

[0016] Estimated value x ｎ The actual error covariance matrix P of the tilde. ｎ As such, the error covariance matrix P ｎ This is the error covariance matrix P output from the radar tracking filter 3. ｎ When it is guaranteed to be less than a tilde (P ｎ Tilde-P ｎ (where is a positive definite matrix), estimated value x １ Tilde, ..., x Ｎ Regardless of the correlation between tildes, the estimated value expressed by the above formula (14) x tilde (ω １ , , ω Ｎ The actual error covariance matrix P of ) ｃｃ (ω １ , , ω Ｎ ) is the error covariance matrix P(ω) calculated by the above formula (15). １ , , ω Ｎ ) It becomes smaller than the tilde (P (ω １ , , ω Ｎ ) Tilde-P ｃｃ (ω １ , , ω Ｎ Non-Patent Literature 1 states that it is guaranteed that the matrix is ​​a positive definite matrix. Non-Patent Literature 1 describes the case where the number of sensors N is "2", but the above point also applies to a general number of sensors N. Furthermore, in the above configuration, the target position is estimated using only the smoothed values ​​from N radars, whereas in the case where estimates are obtained sequentially for each sample, a configuration in which the estimate from the previous sample is used as one of the integrated data for the current sample is also conceivable.

[0017] JP 2008-45910 Simon J. Julier Jeffrey K. Uhlmann, “A Non-divergent Estimation Algorithm in the Presence of Unknown Correlations,” Proceedings of the American Control Conference Albuquerque, New Mexico June 1997

[0018] Conventional signal processing devices use weights ω to integrate multiple target data. １ ,・・・,ω Ｎ There was flexibility in setting the weight ω. １ ,・・・,ω Ｎ Depending on the settings, there was a problem in that the accuracy of estimating the target position decreased.

[0019] This disclosure aims to solve the above-mentioned problems and to provide a signal processing device that can improve the accuracy of target position estimation compared to conventional methods.

[0020] The signal processing device according to this disclosure comprises multiple observation systems that calculate a smoothed value of the target position and an error covariance matrix of the smoothed value using observation data of the target, and an integration system that estimates the target position by integrating multiple smoothed values ​​output from the multiple observation systems using weights set to reduce the error covariance matrix.

[0021] According to this disclosure, the target position is estimated by calculating the smoothed value of the target position and the error covariance matrix of the smoothed value using the target observation data, and integrating multiple smoothed values ​​output from multiple observation systems using weights set to minimize the error covariance matrix. As a result, the signal processing device according to this disclosure can improve the accuracy of target position estimation compared to conventional methods.

[0022] This figure shows an example configuration of a signal processing device according to Embodiment 1. This figure shows an example configuration of a weight setting unit included in the signal processing device according to Embodiment 1. This flowchart shows the signal processing method according to Embodiment 1. This figure shows the status of a conditional optimization problem in Embodiment 1. Figures 5A and 5B show an example of a hardware configuration that realizes the functions of the signal processing device according to Embodiment 1. This figure shows an example configuration of a signal processing device according to Embodiment 2. This figure shows an example configuration of a full-search weight setting unit included in the signal processing device according to Embodiment 2. This flowchart shows the signal processing method according to Embodiment 2. This figure shows an example configuration of a signal processing device according to Embodiment 3. This figure shows an example configuration of a range-limited search weight setting unit included in the signal processing device according to Embodiment 3. This flowchart shows the signal processing method according to Embodiment 3. This figure shows an example configuration of a signal processing device according to Embodiment 4. This figure shows an example configuration of a sensor-selection weight setting unit included in the signal processing device according to Embodiment 4. This flowchart shows the signal processing method according to Embodiment 4. This flowchart shows the procedure for weight calculation in Embodiment 4. This figure shows an example configuration of a conventional signal processing device. This figure shows an example configuration of a radar tracking filter in a conventional signal processing device and in signal processing devices according to Embodiments 1 to 3. This figure shows an example configuration of a data integration unit in a conventional signal processing device and in signal processing devices according to Embodiments 1 to 3.

[0023] Embodiment 1. Figure 1 is a diagram showing an example of the configuration of a signal processing device 1 according to Embodiment 1. The signal processing device 1 is configured to include, for example, a radar 2 which is a sensor, a radar tracking filter 3, a weight setting unit 9, and a data integration unit 10, as shown in Figure 1. In the signal processing device 1, N radar tracking filters 3-#1 to 3-#N constitute a plurality of observation systems, each corresponding to one of the N radars 2-#1 to 2-#N. The weight setting unit 9 and the data integration unit 10 constitute an integration system that estimates the target position by integrating multiple smoothed values ​​output from the plurality of observation systems using weights set to reduce the error covariance matrix. The configuration examples of the radar 2, radar tracking filter 3, and data integration unit 10 in the signal processing device 1 are the same as those of the conventional device described with reference to Figures 16 to 18, so a further explanation is omitted. Hereafter, the target will be assumed to be an aircraft 100.

[0024] Figure 2 shows an example of the configuration of the weight setting unit 9. In Figure 2, the memory circuit 4 stores the observation data transmitted from the radar 2 and the estimated values ​​transmitted from the radar tracking filter 3. The correlation coefficient calculation unit 5 calculates the correlation coefficient between the estimated values ​​output from the radar tracking filters 3 of different radars 2.

[0025] Furthermore, the error covariance matrix calculation unit 6 calculates the error covariance matrix based on the correlation coefficients transmitted from the correlation coefficient calculation unit 5. The evaluation function setting unit 7 finds an evaluation function to integrate the estimated values ​​by reducing the error covariance while guaranteeing an upper limit for the error covariance matrix due to errors in the correlation coefficients. The optimal weight calculation unit 8 calculates the optimal weights using the evaluation function transmitted from the evaluation function setting unit 7. The weight setting unit 9 is composed of a memory circuit 4, a correlation coefficient calculation unit 5, an error covariance matrix calculation unit 6, an evaluation function setting unit 7, and an optimal weight calculation unit 8.

[0026] Next, an example of the operation of the signal processing device 1 according to Embodiment 1 will be described. First, the radar 2-#n observes the direction of existence (azimuth angle) θaz,n and distance Rd,n of the target aircraft 100. Thereafter, the signal processing device 1 operates in the same manner as the conventional device described above, and obtains the estimated value x of the l-th sampling from the radar tracking filter 3. １，ｌTilde, ..., x Ｎ，ｌ Tilde and error covariance matrix P １，１ Tilde, ..., P Ｎ，Ｎ Tilde and are transmitted to the weight setting unit 9 and the data integration unit 10.

[0027] Here, the error covariance matrix P ｎ，ｎ Tilde is the estimated value x ｎ，ｌ Tilde represents the error covariance matrix of the estimated value x. Error covariance matrix P ｎ，ｎ Although the sampling number l is omitted in tilde, it is actually updated for each sampling. In the following equations, the sampling number l is omitted to prevent notation complexity.

[0028] The memory circuit 4 of the weight setting unit 9 stores the observed values z ｄ for a predetermined number of samplings l １，ｌ , ..., z ｎ，ｌ , and the estimated value x １，ｌ Tilde, ..., x Ｎ，ｌ Tilde, and then transmits each stored value to the correlation coefficient calculation unit 5. As a result, the correlation coefficient calculation unit 5 calculates the correlation coefficient using the observation data over the past few samplings. The correlation coefficient calculation unit 5 obtains the correlation coefficient between the estimated values of the radar tracking filters 3 of different radars 2. Specifically, the correlation coefficient calculation unit 5 obtains the correlation coefficient between the estimated values at radar 2-#n and radar 2-#m (1 ≤ n, m ≤ N).

[0029] For example, the correlation coefficient calculation unit 5 calculates the correlation coefficient k ｘｘ，ｎ，ｍ Tilde between the x-axis direction of radar 2-#n and the x-axis direction of radar 2-#m ｙｙ，ｎ，ｍ Tilde between the y-axis direction of radar 2-#n and the y-axis direction of radar 2-#m ｘｙ，ｎ，ｍ Tilde between the x-axis direction of radar 2-#n and the y-axis direction of radar 2-#m ｙｘ，ｎ，ｍ Tilde according to the following equations (17) to (20), respectively. In the following equations (17) to (20), x ｘ，ｎ，ｉ Tilde is the estimated value x ｎ，ｉ The x coordinate of tilde, x ｙ，ｎ，ｉ Tilde is the estimated value x ｎ，ｙThe y-coordinate and z-coordinate of the tilde. ｘ，ｎ，ｉ is the observed value z ｎ，ｉ x coordinate, z ｙ，ｎ，ｉ is the observed value z ｎ，ｉ These represent the y-coordinates of each element.

[0030] Correlation coefficient k xx,n,m Tilde, k yy,n,m Tilde, k xy,n,m Tilde, k yx,n,m The tilde is transmitted from the correlation coefficient calculation unit 5 to the error covariance matrix calculation unit 6. The error covariance matrix calculation unit 6 first calculates the error covariance matrix P between the radars 2. n,m This is calculated using the following formulas (21) to (25). In the following formulas, P n,m(i,k) This is the error covariance matrix P. n,m This represents the element in row i and column k.

[0031] Radar error covariance matrix P ｎ，ｍ The (1 ≤ n, m ≤ N) is transmitted from the error covariance matrix calculation unit 6 to the evaluation function setting unit 7. Also, the error covariance matrix P of the estimated values ​​for each radar 2 is transmitted. １，１ Tilde, ..., P Ｎ，Ｎ The tilde is also transmitted from each radar 2 to the evaluation function setting unit 7.

[0032] The evaluation function setting unit 7 sets the evaluation function h according to the following formulas (26) to (29). ｎ (ω １ , , ω Ｎ Set the evaluation function h. ｎ (ω １ , , ω Ｎ This is transmitted from the evaluation function setting unit 7 to the optimal weight calculation unit 8. In the following equation (26), Λ represents a function that calculates the diagonal sum of a matrix.

[0033] The optimal weight calculation unit 8 calculates the weights that minimize the error covariance. Figure 3 is a flowchart showing the signal processing method according to Embodiment 1, illustrating the processing procedure for weight calculation in the optimal weight calculation unit 8. This processing procedure is similar to that of the Newton method.

[0034] First, in procedure ST1-1, the optimal weight calculation unit 8 generates a vector ω = [ω] representing the initial weight values. １ , , ω Ｎ ] Ｔ The optimal weight calculation unit 8 sets the evaluation function h transmitted from the evaluation function setting unit 7. ｎ (ω １ , , ω Ｎ Find the derivative of ). Equations (30) to (35) below are the evaluation function h ｎ (ω １ , , ω Ｎ This represents the derivative of ).

[0035] In procedure ST1-2, the optimal weight calculation unit 8 calculates the evaluation function h obtained in procedure ST1-1. ｎ (ω １ , , ω Ｎ The matrix of the following equation (36), whose components are the derivatives of ), is obtained. In addition, the optimal weight calculation unit 8 calculates the evaluation function h ｎ (ω １ , , ω Ｎ Find the following column vector (37) whose components are ).

[0036] In procedure ST1-3, the optimal weight calculation unit 8 updates the weight vector ω using the Newton method, and the correction vector Δ = φ －１ To find h. In procedure ST1-4, the optimal weight calculation unit 8 changes the vector ω - Δ to the weight vector ω. From the above equation (16), the weight ω ｎ The range of (n=1, ..., N) is 0 ≤ ω ｎIf the value is ≤ 1 and there is a negative component in the newly set weight vector ω, the optimal weight calculation unit 8 corrects that component to zero in step ST1-5.

[0037] Furthermore, since the sum of the weights is "1" according to the above formula (16), in step ST1-6, the optimal weight calculation unit 8 modifies the weight vector ω to the following formula (38) so that the sum of the weights becomes 1.

[0038] In procedure ST1-7, the optimal weight calculation unit 8 performs a convergence determination. Criteria for determining convergence include when the norm of the correction vector Δ becomes sufficiently small, and when a predetermined number of updates have been performed. If convergence is determined, the optimal weight calculation unit 8 calculates the optimal weight ω １ , , ω Ｎ The output is displayed. If it is determined that the result has not converged, the process proceeds to step ST1-8.

[0039] In procedure ST1-8, the optimal weight calculation unit 8 calculates an evaluation function h related to the updated weight vector ω. ｎ (ω １ , , ω Ｎ ) and its derivative are calculated. Evaluation function h ｎ (ω １ , , ω Ｎ The result and its derivative are transmitted to step ST1-2. The process is then repeated until it is determined that convergence has occurred in step ST1-7. The optimal weight ω is output when convergence is determined to have occurred. １ , , ω Ｎ This is transmitted from the optimal weight calculation unit 8 to the data integration unit 10. Thereafter, the signal processing device 1 operates in the same manner as conventional devices, and the smoothing values ​​transmitted from the radar tracking filter 3 are integrated by the data integration unit 10.

[0040] Here, the evaluation function h ｎ (ω １ , , ω Ｎ We will explain how the optimal weights can be calculated by performing the series of operations shown in Figure 3 using the evaluation function h. ｎ (ω １ , , ω Ｎ) is the weight ω for which the diagonal sum of the error covariances shown in equation (39) below for the estimated value x(ω) tilde in equation (14) above is minimized under the conditions of equation (16) above. １ , , ω Ｎ This represents the necessary conditions.

[0041] The error covariance shown in equation (39) above is P ｃｃ If we let (ω), then P ｃｃ (ω) can be expressed by the following equations (40) and (41).

[0042] P ｃｃ The diagonal sum of (ω) Λ(P) ｃｃ The weight ω that minimizes (ω) １ , , ω Ｎ The constraint condition ω in equation (16) above １ +...+ω Ｎ By solving the conditional optimization problem under the condition = 1, we obtain the optimal weight ω １ , , ω Ｎ Here, the weight ω is found. １ ω １ +Δ １ , weight ω ｎ ω ｎ +Δ ｎ When only this is changed, the following equation (42) is obtained.

[0043] Figure 4 shows the state of the conditional optimization problem at this time. Using a similar approach to the Lagrange multiplier method, which is a well-known method for solving conditional optimization problems, Λ(P ｃｃ (ω １ , , ω Ｎ )) = constant weight ω １ , , ω Ｎ The region (solid line in the figure) and the constraint condition ω of the above equation (16) １ +...+ω Ｎ The region where = 1 (dotted line in the figure) is tangent at the optimal point, as shown in Figure 4. Therefore, in the vicinity of the optimal point, Λ(P ｃｃ (ω １ , , ω Ｎ )) = constant, weight ω１ A minute amount Δ １ , weight ω ｎ A minute amount Δ ｎ The slope Δ when only that much is changed ｎ / Δ １ is a constraint ω １ +...+ω Ｎ Under the condition = 1, weight ω １ A minute amount Δ １ , weight ω ｎ A minute amount Δ ｎ When only this much is changed, the slope becomes equal to -1. The parts of both sides of equation (42) above that are expressed by equation (43) below cancel each other out, and Δ ｎ / Δ １ = 1.

[0044] Paying attention to the above relationship, equation (42) can be expressed as equation (44) below.

[0045] Therefore, the left side of equation (44) above is the evaluation function h ｎ (ω １ , , ω Ｎ By doing so, the above equation (26) is calculated. ｎ (ω １ , , ω Ｎ The fact that ) = 0 means that the optimal weight ω １ , , ω Ｎ It is a necessary condition, and there are other constraints ω １ +...+ω Ｎ = 1 is also required. Constraint ω １ +...+ω Ｎ = 1 always occurs in the process shown in Figure 3, ω １ +...+ω Ｎ The search is satisfied by keeping the condition = 1. In fact, by setting all elements in the first row of matrix Φ to "1" and the values ​​in the first row of vector h to "0", the sum of each component of the change amount Δ of the weight vector ω is set to "0". Therefore, ω is set as the initial value. １ +...+ω Ｎ By setting a weight vector ω such that = 1, we can always ω １ +...+ω Ｎ The weight vector ω is updated so that it equals 1.

[0046] As described above, the signal processing device 1 according to Embodiment 1 is configured to reduce the error covariance and calculate weights for integrating smoothed values, thereby improving the accuracy of the estimated value obtained by integrating the smoothed values ​​compared to conventional methods. The influence of the error in the correlation estimate between the smoothed values ​​of each sensor on the error of the integrated estimated value is guaranteed to have an upper limit, and the signal processing device 1 can perform robust estimation with respect to the error in the correlation estimate. Furthermore, the optimal weight calculation unit 8 of the signal processing device 1 uses the Newton method, enabling high-speed calculation of the optimal weight. Here, an example using radar 2 as the sensor has been described, but other sensors can be used as the sensor, not just radar 2. Moreover, the signal processing device 1 can also be configured using a method that employs time-difference positioning and GPS positioning to estimate the position of the aircraft 100.

[0047] Next, with reference to Figure 5, an example of a hardware configuration that realizes the functions of the signal processing device 1 according to Embodiment 1 will be described. The functions of the radar tracking filter 3 (coordinate transformation unit 18, Kalman gain calculation unit 19, smoothing value calculation unit 20, and predicted value calculation unit 21), the weight setting unit 9 (memory circuit 4, correlation coefficient calculation unit 5, error covariance matrix calculation unit 6, evaluation function setting unit 7, and optimal weight calculation unit 8), and the data integration unit 10 (smooth value memory circuit 22, covariance memory circuit 23, and estimated value calculation unit 24) provided by the signal processing device 1 are realized by a processing circuit. The processing circuit may be dedicated hardware as shown in Figure 5A, or it may be a CPU (Central Processing Unit, central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, or DSP (Digital Signal Processor)) 52 that executes the program stored in the memory 53, as shown in Figure 5B.

[0048] If the processing circuit is dedicated hardware, the processing circuit 51 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination thereof. The functions of the radar tracking filter 3, the weight setting unit 9, and the data integration unit 10 may be implemented by separate processing circuits 51, or they may be implemented together by a single processing circuit 51.

[0049] When the processing circuit is a CPU 52, the functions of the radar tracking filter 3, the weight setting unit 9, and the data integration unit 10 are realized by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in memory 53. The CPU 52 realizes the functions of each unit by reading and executing the programs recorded in memory 53. That is, the signal processing device 1 has memory for storing programs that, when executed by the processing circuit, result in the execution of each step shown in Figure 3, for example. These programs also cause the computer to execute the procedures and methods of the radar tracking filter 3, the weight setting unit 9, and the data integration unit 10. Here, memory 53 includes, for example, non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable ROM), EEPROM (Electrically EPROM), magnetic disks, flexible disks, optical disks, compact disks, minidiscs, or DVDs (Digital Versatile Discs).

[0050] Furthermore, the functions of the radar tracking filter 3, weight setting unit 9, and data integration unit 10 may be partially implemented by dedicated hardware and partially by software or firmware. For example, the radar tracking filter 3 can be implemented by a processing circuit as dedicated hardware, while the weight setting unit 9 and data integration unit 10 can be implemented by a processing circuit reading and executing a program stored in memory 53.

[0051] Thus, the processing circuit can realize each of the above-mentioned functions through hardware, software, firmware, or a combination thereof.

[0052] As described above, the signal processing device 1 according to Embodiment 1 includes a plurality of observation systems that calculate a smoothed value of the aircraft 100's position and an error covariance matrix of the smoothed value using observation data of the aircraft 100, and an integration system that estimates the position of the aircraft 100 by integrating a plurality of smoothed values ​​output from the plurality of observation systems using weights set to reduce the error covariance matrix. As a result, the signal processing device 1 can improve the accuracy of estimating the position of the aircraft 100 compared to conventional systems.

[0053] In the signal processing device 1 according to Embodiment 1, the N observation systems each include radar tracking filters 3-#1 to 3-#N that calculate smoothed values ​​using observation data including the direction of presence of the aircraft 100 and the distance to the aircraft 100 acquired by each of the radars 2-#1 to 2-#N, and output smoothed values ​​and error covariance matrices. The integration system includes a weight setting unit 9 that sets weights for integrating the smoothed values ​​while suppressing the increase in the error covariance matrices output from the radar tracking filters 3-#1 to 3-#N, and a data integration unit 10 that integrates the smoothed values ​​using the set weights and estimates the position of the aircraft 100 based on the obtained estimated values. As a result, the signal processing device 1 can improve the accuracy of estimating the position of the aircraft 100 compared to conventional systems.

[0054] In the signal processing device 1 according to Embodiment 1, the radar tracking filter 3 includes a coordinate transformation unit 18 that transforms the coordinate system of the aircraft 100's observation data from a polar coordinate system to a Cartesian coordinate system, a Kalman gain calculation unit 19 that calculates the Kalman gain used when calculating the smoothing value in the tracking process, a smoothing value calculation unit 20 that calculates the smoothing value using the observation data transformed by the coordinate transformation unit 18 and the Kalman gain calculated by the Kalman gain calculation unit 19, and a prediction value calculation unit 21 that predicts the position of the aircraft 100 in the next sampling. As a result, the signal processing device 1 can accurately calculate the smoothing value and the error covariance matrix of the smoothing value.

[0055] In the signal processing device 1 according to Embodiment 1, the predicted value calculation unit 21 predicts the smoothed value and error covariance matrix for the next sampling, the Kalman gain calculation unit 19 calculates the Kalman gain using the error covariance matrix for the next sampling predicted by the predicted value calculation unit 21, and the smoothed value calculation unit 20 calculates the error covariance matrix for the current sampling using the Kalman gain calculated by the Kalman gain calculation unit 19 and the error covariance matrix for the next sampling predicted by the predicted value calculation unit 21. As a result, the signal processing device 1 can accurately calculate the smoothed value and the error covariance matrix of the smoothed value.

[0056] In the signal processing device 1 according to Embodiment 1, the data integration unit 10 includes a smoothing value memory circuit 22 for storing smoothing values ​​transmitted from the radar tracking filter 3, a covariance memory circuit 23 for storing the error covariance matrix transmitted from the radar tracking filter 3, and an estimated value calculation unit 24 that integrates a plurality of smoothing values ​​based on the smoothing values ​​stored in the smoothing value memory circuit 22 and the error covariance matrix stored in the covariance memory circuit 23 to calculate an estimated value of the aircraft 100's position. As a result, the signal processing device 1 can calculate the estimated value with high accuracy.

[0057] In the signal processing device 1 according to Embodiment 1, the weight setting unit 9 includes a memory circuit 4 that stores observation data transmitted from the radar 2 and smoothed values ​​transmitted from the radar tracking filter 3, a correlation coefficient calculation unit 5 that calculates the correlation coefficient between smoothed values ​​output from radar tracking filters 3 corresponding to different radars 2, an error covariance matrix calculation unit 6 that calculates an error covariance matrix based on the correlation coefficient transmitted from the correlation coefficient calculation unit 5, an evaluation function setting unit 7 that calculates an evaluation function for integrating smoothed values ​​while suppressing the increase in the error covariance matrix and guaranteeing an upper limit of the error covariance matrix due to errors in the correlation coefficient, and an optimal weight calculation unit 8 that calculates weights using the evaluation function transmitted from the evaluation function setting unit 7. As a result, the signal processing device 1 can accurately calculate the optimal weights.

[0058] In the signal processing device 1 according to Embodiment 1, the optimal weight calculation unit 8 calculates the optimal weight by applying the Newton method using an evaluation function. As a result, the signal processing device 1 can calculate the optimal weight at high speed.

[0059] In the signal processing device 1 according to Embodiment 1, the correlation coefficient calculation unit 5 calculates the correlation coefficient using observation data spanning several past samples. As a result, the signal processing device 1 can calculate the correlation coefficient with high accuracy.

[0060] In the signal processing device 1 according to Embodiment 1, the optimal weight calculation unit 8 sets any weights with a negative sign to 0 when calculating the optimal weights, and further adjusts the weights so that the sum of the weights becomes 1. As a result, the signal processing device 1 can efficiently calculate the optimal weights.

[0061] The signal processing method according to Embodiment 1 includes the steps of: a radar tracking filter 3 corresponding to each of the multiple radars 2 calculates a smoothed value indicating the smoothed position of the aircraft 100 using observation data acquired by each radar 2, including the direction of the aircraft 100's presence and the distance to the target, and outputs the calculated smoothed value and the error covariance matrix of the smoothed value; a weight setting unit 9 sets optimal weights for integrating the smoothed values ​​while suppressing the increase in the output error covariance matrix of the smoothed value; and a data integration unit 10 integrates the smoothed values ​​using the set optimal weights and estimates the position of the aircraft 100 based on the obtained estimated value. By the signal processing device 1 executing the above method, the accuracy of estimating the position of the aircraft 100 can be improved compared to conventional methods.

[0062] The program according to Embodiment 1 causes the computer to function as a plurality of radar tracking filters 3, each corresponding to a plurality of radars 2, which calculate a smoothed value indicating the smoothed position of the aircraft 100 using observation data acquired by each radar 2, including the direction of the aircraft 100's presence and the distance to the aircraft 100, and output the calculated smoothed value and the error covariance matrix of the smoothed value; a weight setting unit 9 that sets weights for integrating the smoothed values ​​while suppressing the increase of the output error covariance matrix; and a data integration unit 10 that integrates the smoothed values ​​using the set weights and estimates the position of the aircraft 100 based on the obtained estimated value. By having the computer execute the above program, a signal processing device 1 can be provided that has improved accuracy in estimating the position of the aircraft 100 compared to the conventional method.

[0063] Embodiment 2. Embodiment 1 described a signal processing device that can improve the accuracy of the estimated value obtained by integrating smoothed values ​​by calculating weights that reduce the error covariance and integrating smoothed values ​​compared to conventional methods. Embodiment 2 describes a signal processing device that, in addition to the above, can prevent the calculation of suboptimal weights such as local optima.

[0064] Figure 6 shows an example of the configuration of the signal processing device 1b according to Embodiment 2. The signal processing device 1b differs from the signal processing device 1 shown in Figure 1 in that the weight setting unit 9 has been changed to a full-search type weight setting unit 13. The other components of the signal processing device 1b according to Embodiment 2 are the same as those of the signal processing device 1 shown in Figure 1, so the same reference numerals are used and their descriptions are omitted.

[0065] Figure 7 shows an example of the configuration of the exhaustive search type weight setting unit 13. In Figure 7, the memory circuit 4, the correlation coefficient calculation unit 5, and the error covariance matrix calculation unit 6 are the same as the memory circuit 4, the correlation coefficient calculation unit 5, and the error covariance matrix calculation unit 6 provided in the weight setting unit 9 in Embodiment 1.

[0066] Furthermore, in Figure 7, the search-type evaluation function setting unit 11 searches for evaluation values ​​to reduce the error covariance and integrate the estimated values. The exhaustive search-type optimal weight calculation unit 12 calculates the optimal weights by exhaustive search using the evaluation function transmitted from the search-type evaluation function setting unit 11. The exhaustive search-type weight setting unit 13 is composed of a memory circuit 4, a correlation coefficient calculation unit 5, an error covariance matrix calculation unit 6, a search-type evaluation function setting unit 11, and a exhaustive search-type optimal weight calculation unit 12.

[0067] Next, an example of the operation of the signal processing device 1b will be described. Radar 2-#n is the azimuth angle θ, which is the direction of the target aircraft 100. ａｚ，ｎ distance R ｄ，ｎ The system observes the following. Subsequently, the signal processing device 1b operates in the same manner as the signal processing device 1, and estimates x from the radar tracking filter 3. １，ｌ Tilde, ..., x Ｎ，ｌ The tilde and the error covariance matrix P １，１ Tilde, ..., P Ｎ，Ｎ The tilde and data are transmitted to the data integration unit 10. Also, the observed value z is received from radar 2-#n. １，ｌ , ..., z ｎ，ｌ However, the estimated value x from the radar tracking filter 3 １，ｌ Tilde, ..., x Ｎ，ｌ The tilde and the error covariance matrix P １，１ Tilde, ..., P Ｎ，ＮEach tilde is transmitted to the exhaustive search weight setting unit 13. Thereafter, the signal processing unit 1b operates in the same manner as the signal processing unit 1, and the error covariance matrix P between radars is generated. ｎ，ｍ The values ​​(1 ≤ n, m ≤ N) are transmitted from the error covariance matrix calculation unit 6 to the search-type evaluation function setting unit 11.

[0068] The search-type evaluation function setting unit 11 searches for an evaluation value to reduce the error covariance and integrate the estimated values. Specifically, the search-type evaluation function setting unit 11 uses the following equation (45) to determine the evaluation function I(ω １ , , ω Ｎ Set ). In the following equation (45), P ｃｃ (ω １ , , ω Ｎ The tilde represents the weighted error covariance matrix defined by equation (15) above, and Q ｃｃ (ω １ , , ω Ｎ ) represents the matrices defined by equation (27) above. Note that Λ represents a function that calculates the diagonal sum of the matrices.

[0069] The evaluation function I(ω) set by the search-type evaluation function setting unit 11 １ , , ω Ｎ This information is transmitted from the search-type evaluation function setting unit 11 to the exhaustive search-type optimal weight calculation unit 12.

[0070] Figure 8 is a flowchart showing the signal processing method according to Embodiment 2, and represents the processing procedure for weight calculation by the exhaustive search type optimal weight calculation unit 12. First, in procedure ST2-1, the exhaustive search type optimal weight calculation unit 12 sets various initial values. For example, the exhaustive search type optimal weight calculation unit 12 sets the number of divisions Nω in the range of 0 to 1 that the weight can take. Next, the exhaustive search type optimal weight calculation unit 12 sets the type of weight according to the following equations (46) and (47). In the following equation (46), ω ｋ，ｎΔｋ (1 ≤ k ≤ N) is the weight ω ｋ This represents the type of weight set for that.

[0071] Next, in step ST2-2, the exhaustive search type optimal weight calculation unit 12 calculates that the sum of all weight types is 1 (ω １，ｎΔ１ +ω ２，ｎΔ２ +...+ω Ｎ，ｎΔＮ We examine all combinations such that = 1). There are N such combinations. ｃｏｍｂ Assuming there are n ｃ (1 ≤ n) ｃ ≤N ｃｏｍｂ The th-th combination of weights found can be expressed by the following equation (48).

[0072] In procedure ST2-3, the exhaustive search type optimal weight calculation unit 12 uses an initial value n ｃ Set = 1. In procedure ST2-4, the exhaustive search type optimal weight calculation unit 12 evaluates the function I(ω(n) using the following equations (49) and (50). ｃ Calculate )).

[0073] In procedure ST2-5, the exhaustive search type optimal weight calculation unit 12 calculates the number n ｃ = N ｃｏｍｂ The system determines whether this is the case, and if it is, proceeds to step ST2-7. Otherwise, it proceeds to step ST2-6, and the exhaustive search type optimal weight calculation unit 12 calculates the number n. ｃ to n ｃ Add +1 and proceed to step ST2-4.

[0074] In procedure ST2-7, the exhaustive search-type optimal weight calculation unit 12 minimizes the evaluation function to the number n according to the following equation (51). ｃ，ｍｉｎ Investigate.

[0075] In procedure ST2-8, the exhaustive search type optimal weight calculation unit 12 calculates the evaluation function I(ω(n)) according to the following formula (52). ｃ Calculate the weights (optimal weights) that minimize ).

[0076] Optimal weight ω １ , , ω ＮThis is transmitted from the exhaustive search-type optimal weight calculation unit 12 to the data integration unit 10. Thereafter, the signal processing device 1b operates in the same manner as the signal processing device 1, and the smoothed values ​​transmitted from the radar tracking filter 3 are integrated in the data integration unit 10.

[0077] As the signal processing device 1b is configured as described above, it can calculate weights that integrate smoothed values ​​while minimizing the error covariance, thereby improving the accuracy of the estimated value obtained by integrating the smoothed values ​​compared to conventional methods. Furthermore, the signal processing device 1b calculates the optimal weights by exhaustive search, preventing the calculation of suboptimal weights such as local optima.

[0078] As described above, the signal processing device 1b according to Embodiment 2 includes, instead of the evaluation function setting unit 7, a search-type evaluation function setting unit 11 that finds an evaluation function for suppressing the increase in error covariance and integrating smoothed values ​​by searching, and instead of the optimal weight calculation unit 8, a full-search type optimal weight calculation unit 12 that calculates the optimal weights by performing a full search using the evaluation function transmitted from the search-type evaluation function setting unit 11. As a result, the signal processing device 1b can prevent the calculation of suboptimal weights such as local optima.

[0079] Embodiment 3. Embodiment 2 described a signal processing device that calculates the optimal weight by exhaustive search. Embodiment 3 describes a processing device that can reduce the processing load required for the search by limiting the search range for the optimal weight.

[0080] Figure 9 shows an example of the configuration of the signal processing device 1c according to Embodiment 3. The signal processing device 1c differs from the signal processing device 1b shown in Figure 5 in that the exhaustive search type weight setting unit 13 has been changed to a range-limited search type weight setting unit 15. The other components of the signal processing device 1c are the same as those of the signal processing device 1b shown in Figure 5, so the same reference numerals are used and their descriptions are omitted.

[0081] Figure 10 shows an example of the configuration of the range-limited search type weight setting unit 15. In Figure 10, the memory circuit 4, correlation coefficient calculation unit 5, error covariance matrix calculation unit 6, and search type evaluation function setting unit 11 are the same as the memory circuit 4, correlation coefficient calculation unit 5, error covariance matrix calculation unit 6, and search type evaluation function setting unit 11 provided in the full search type weight setting unit 13 in Embodiment 2.

[0082] Furthermore, in Figure 10, the range-limited search type optimal weight calculation unit 14 performs a range-limited search using the evaluation function transmitted from the search type evaluation function setting unit 11 and calculates the optimal weights. The range-limited search type weight setting unit 15 is composed of a memory circuit 4, a correlation coefficient calculation unit 5, an error covariance matrix calculation unit 6, a search type evaluation function setting unit 11, and the range-limited search type optimal weight calculation unit 14.

[0083] Next, an example of the operation of the signal processing device 1c will be described. Radar 2-#n is the azimuth angle θ, which is the direction of the target aircraft 100. ａｚ，ｎ distance R ｄ，ｎ The signal processing device 1c operates in the same manner as the signal processing device 1, and estimates x from the radar tracking filter 3. １，ｌ Tilde, ..., x Ｎ，ｌ The tilde and the error covariance matrix P １，１ Tilde, ..., P Ｎ，Ｎ The tilde and data are transmitted to the data integration unit 10. Also, the observed value z is received from radar 2-#n. １，ｌ , ..., z ｎ，ｌ However, the estimated value x from the radar tracking filter 3 １，ｌ Tilde, ..., x Ｎ，ｌ The tilde and the error covariance matrix P １，１ Tilde, ..., P Ｎ，Ｎ Each tilde is transmitted to the range-limited search type weight setting unit 15. Thereafter, the signal processing unit 1c operates in the same manner as the signal processing unit 1b, and the evaluation function I(ω １ , , ω Ｎ This information is transmitted from the search-type evaluation function setting unit 11 to the range-limited search-type optimal weight calculation unit 14.

[0084] Figure 11 is a flowchart showing the signal processing method according to Embodiment 3, illustrating the processing procedure for weight calculation by the range-limited search type optimal weight calculation unit 14. First, in procedure ST3-1, the range-limited search type optimal weight calculation unit 14 sets various initial values. For example, the range-limited search type optimal weight calculation unit 14, similar to the full-search type optimal weight calculation unit 12 in Embodiment 2, sets the number of divisions N in the range of 0 to 1 that the weight can take. ω Set it.

[0085] Next, the range-limited search type optimal weight calculation unit 14 sets the type of weight according to the above equations (46) and (47).

[0086] Furthermore, the range-limited search type optimal weight calculation unit 14 determines the center ω of the search range. ｎ The tilde (1 ≤ n ≤ N) and the width of the search range Δ ωｎ (1 ≤ n ≤ N) is set. Specifically, the range-limited search type optimal weight calculation unit 14 sets the center of the search range ω ｎ Tilde and search range width Δ ωｎ The optimal weight is set within the range in which an optimal weight is expected to exist. For example, when tracking an aircraft 100, the optimal weight is calculated for each sample, so the range-limited search type optimal weight calculation unit 14 sets the optimal weight calculated in the previous sample as the center of the search range, and the width of the search range can be set within a range in which the processing load of the search can be sufficiently reduced.

[0087] In procedure ST3-2, the range-limited search type optimal weight calculation unit 14 determines that the sum of all weight types is 1 (ω １，ｎΔ１ +ω ２，ｎΔ２ +...+ω Ｎ，ｎΔＮ Find the combinations that satisfy = 1) within the range of the following formula (53).

[0088] N such combinations ｃｏｍｂ Assuming there are n, the range-limited search type optimal weight calculation unit 14 calculates n ｃ (1 ≤ n) ｃ ≤N ｃｏｍｂThe )th combination of weights found is represented by the above formula (48). Subsequent steps ST3-3 to ST3-8 are the same as steps ST2-3 to ST2-8 in Embodiment 2. The optimal weight ω calculated in step ST3-8 １ , , ω Ｎ The data is transmitted from the range-limited search type optimal weight calculation unit 14 to the data integration unit 10, where the smoothed values ​​transmitted from the radar tracking filter 3 are integrated.

[0089] The signal processing device 1c according to Embodiment 3 is configured as described above, so that it can calculate weights that integrate smoothed values ​​while reducing the error covariance, and improve the accuracy of the estimated value with integrated smoothed values ​​compared to conventional methods. In addition, the signal processing device 1c calculates the optimal weights by limiting the search range, which reduces the processing load compared to when the optimal weights are calculated by exhaustive search. Furthermore, the signal processing device 1c can prevent local optima from being searched and resulting in the calculation of suboptimal weights.

[0090] As described above, the signal processing device 1c according to Embodiment 3 includes a range-limited search type optimal weight calculation unit 14 that calculates the optimal weight by performing a range-limited search using an evaluation function transmitted from the search-type evaluation function setting unit 11, instead of the optimal weight calculation unit 8. As a result, the signal processing device 1c can reduce the processing load required for the search, in addition to the effects of Embodiment 1.

[0091] In the signal processing device 1c according to Embodiment 3, the range-limited search type optimal weight calculation unit 14 searches for the optimal weight for the current sampling only in the vicinity of the optimal weight used in the previous sampling. This makes it possible for the signal processing device 1c to reduce the processing load required for the search.

[0092] Embodiment 4. Embodiment 3 described a processing device that can reduce the processing load required for searching by limiting the search range for the optimal weight. Embodiment 4 describes a processing device that can improve the accuracy of the estimated value obtained by integrating the smoothed values ​​by calculating weights using only the smoothed values ​​of valid sensors whose sign is not negative.

[0093] Figure 12 shows an example of the configuration of a signal processing device 1d according to Embodiment 4. The signal processing device 1d differs from the signal processing device 1 shown in Figure 1 in that the weight setting unit 9 has been changed to a sensor-selection type weight setting unit 17. The other components of the signal processing device 1d are the same as those of the signal processing device 1, so they are given the same reference numerals and their descriptions are omitted.

[0094] Figure 13 shows an example of the configuration of the sensor-selection type weight setting unit 17. In Figure 13, the memory circuit 4, correlation coefficient calculation unit 5, error covariance matrix calculation unit 6, and evaluation function setting unit 7 are the same as the memory circuit 4, correlation coefficient calculation unit 5, error covariance matrix calculation unit 6, and evaluation function setting unit 7 provided in the weight setting unit 9 in Embodiment 1.

[0095] Furthermore, in Figure 13, the sensor-selection type optimal weight calculation unit 16 calculates the optimal weights by excluding sensors with negative weights. The sensor-selection type weight setting unit 17 is composed of a memory circuit 4, a correlation coefficient calculation unit 5, an error covariance matrix calculation unit 6, an evaluation function setting unit 7, and the sensor-selection type optimal weight calculation unit 16.

[0096] Next, an example of the operation of the signal processing device 1d will be described. First, radar 2-#n is set to the direction of presence (azimuth angle) θ of the target aircraft 100. ａｚ，ｎ distance R ｄ，ｎ The system observes the following. Subsequently, the signal processing device 1d operates in the same manner as the signal processing device 1, and estimates x from the radar tracking filter 3. １，ｌ Tilde, ..., x Ｎ，ｌ The tilde and the error covariance matrix P １，１ Tilde, ..., P Ｎ，Ｎ The tilde and data are transmitted to the data integration unit 10. Also, the observed value z is received from radar 2-#n. １，ｌ , ..., z ｎ，ｌ However, the estimated value x from the radar tracking filter 3 １，ｌ Tilde, ..., x Ｎ，ｌ The tilde and the error covariance matrix P １，１ Tilde, ..., P Ｎ，Ｎ The tilde is transmitted to the sensor-selective weight setting unit 17. The signal processing unit 1d operates in the same manner as the signal processing unit 1, and the evaluation function h(ω １, , ω Ｎ This information is transmitted from the evaluation function setting unit 7 to the sensor-selection type optimal weight calculation unit 16.

[0097] Figure 14 is a flowchart showing the signal processing method according to Embodiment 4, and illustrates the processing procedure for weight calculation by the sensor-selection type optimal weight calculation unit 16. First, in procedure ST4-1, the sensor-selection type optimal weight calculation unit 16 calculates the initial weight vector ω = [ω １ ...ω Ｎ ] Ｔ Next, the sensor-selection type optimal weight calculation unit 16 receives the evaluation function h transmitted from the evaluation function setting unit 7. ｎ (ω １ , , ω Ｎ The derivative of ) is calculated. The subsequent steps ST4-2 to ST4-4 are the same as steps ST1-2 to ST1-4 in Embodiment 1.

[0098] In procedure ST4-5, the sensor-selection type optimal weight calculation unit 16 checks whether or not there is a component with a negative sign in the weight ω.

[0099] Figure 15 is a flowchart showing the weight calculation procedure in Embodiment 4, and shows a series of processes performed between A and B in Figure 14. If there is a weight ω with a negative sign, the sensor-selection type optimal weight calculation unit 16 proceeds to procedure ST4-9 shown in Figure 15. In procedure ST4-9, the sensor-selection type optimal weight calculation unit 16 excludes the sensors corresponding to the weights with a negative sign. In procedure ST4-10, the sensor-selection type optimal weight calculation unit 16 selects only the sensors corresponding to the weights with a positive sign and resets the sensor type and the number of sensors N. In procedure ST4-11, the sensor-selection type optimal weight calculation unit 16 sets the vector ω = [ω] representing the initial weight values. １ ...ω Ｎ ] Ｔ The sensor-selection type optimal weight calculation unit 16 uses the above equations (30) to (35) to calculate the evaluation function h transmitted from the evaluation function setting unit 7. ｎ (ω １ , , ω ＮThe derivative of ) is also calculated. If there is no weight ω with a negative sign, proceed to step ST4-6. The processing in steps ST4-6 to ST4-8 is the same as in steps ST1-6 to ST1-8 in Embodiment 1. From here on, the sensor-selection type optimal weight calculation unit 16 operates in the same manner as in Embodiment 1, and the optimal weight ω １ , , ω Ｎ Outputs.

[0100] In this way, the sensor-selection type optimal weight calculation unit 16 calculates the weights that minimize the error covariance. Figures 14 and 15 show the processing procedure for weight calculation similar to that of the Newton method.

[0101] As described above, the signal processing device 1d according to Embodiment 4 is configured to calculate weights using only the smoothed values ​​of effective sensors whose weights do not have a negative sign, thereby improving the accuracy of the estimated value obtained by integrating the smoothed values.

[0102] As described above, the signal processing device 1d according to Embodiment 4 includes a sensor-selection type optimal weight calculation unit 16 that calculates weights using only the smoothed values ​​of effective sensors whose weights do not have a negative sign, instead of the optimal weight calculation unit 8. As a result, the signal processing device 1d can improve the accuracy of the estimated values ​​in addition to the effects of Embodiment 1.

[0103] The signal processing device relating to this disclosure can be used, for example, in a radar system.

[0104] Furthermore, this disclosure allows for free combination of each embodiment, modification of any component of each embodiment, or omission of any component in each embodiment.

[0105] 1, 1b, 1c, 1d Signal processing device, 2 Radar, 3 Tracking filter for radar, 4 Memory circuit, 5 Correlation coefficient calculation unit, 6 Error covariance matrix calculation unit, 7 Evaluation function setting unit, 8 Optimal weight calculation unit, 9 Weight setting unit, 10 Data integration unit, 11 Search-type evaluation function setting unit, 12 Exhaustive search-type optimal weight calculation unit, 13 Exhaustive search-type weight setting unit, 14 Range-limited search-type optimal weight calculation unit, 15 Range-limited search-type weight setting unit, 16 Sensor-selection-type optimal weight calculation unit, 17 Sensor-selection-type weight setting unit, 18 Coordinate transformation unit, 19 Kalman gain calculation unit, 20 Smoothed value calculation unit, 21 Predicted value calculation unit, 22 Smoothed value memory circuit, 23 Covariance memory circuit, 24 Estimated value calculation unit, 51 Processing circuit, 52 CPU, 53 Memory, 100 Aircraft.

Claims

1. A signal processing device comprising: multiple observation systems that calculate a smoothed value of a target position and an error covariance matrix of the smoothed value using target observation data; and an integration system that estimates the target position by integrating multiple smoothed values ​​output from the multiple observation systems using weights set to minimize the error covariance matrix.

2. The signal processing apparatus according to claim 1, wherein each of the multiple observation systems includes a radar tracking filter that calculates the smoothed value using observation data including the direction of existence of the target and the distance to the target, acquired by each of the multiple radars, and outputs the calculated smoothed value and the error covariance matrix, and the integration system includes a weight setting unit that sets weights for integrating the smoothed values ​​while suppressing the increase of the error covariance matrix output from the radar tracking filter, and a data integration unit that integrates the smoothed values ​​using the set weights and estimates the target position based on the obtained estimated value.

3. The signal processing apparatus according to claim 2, wherein the radar tracking filter comprises: a coordinate transformation unit that transforms the coordinate system of the target observation data from a polar coordinate system to a Cartesian coordinate system; a Kalman gain calculation unit that calculates a Kalman gain used when calculating the smoothing value in the tracking process; a smoothing value calculation unit that calculates the smoothing value using the observation data transformed by the coordinate transformation unit and the Kalman gain calculated by the Kalman gain calculation unit; and a prediction value calculation unit that predicts the target position in the next sampling.

4. The signal processing apparatus according to claim 3, characterized in that the prediction value calculation unit predicts the smoothed value and the error covariance matrix for the next sampling; the Kalman gain calculation unit calculates the Kalman gain using the error covariance matrix for the next sampling predicted by the prediction value calculation unit; and the smoothed value calculation unit calculates the error covariance matrix for the current sampling using the Kalman gain calculated by the Kalman gain calculation unit and the error covariance matrix for the next sampling predicted by the prediction value calculation unit.

5. The signal processing apparatus according to any one of claims 2 to 4, wherein the data integration unit comprises: a smoothing value memory circuit for storing the smoothing value transmitted from the radar tracking filter; a covariance memory circuit for storing the error covariance matrix transmitted from the radar tracking filter; and an estimated value calculation unit for integrating a plurality of smoothing values ​​and calculating an estimated value of the target position based on the smoothing value stored in the smoothing value memory circuit and the error covariance matrix stored in the covariance memory circuit.

6. The signal processing apparatus according to any one of claims 2 to 5, wherein the weight setting unit comprises: a memory circuit for storing observation data transmitted from the radar and the smoothing values ​​transmitted from the radar tracking filter; a correlation coefficient calculation unit for determining the correlation coefficient between the smoothing values ​​output from the radar tracking filters corresponding to different radars; an error covariance matrix calculation unit for calculating the error covariance matrix based on the correlation coefficient transmitted from the correlation coefficient calculation unit; an evaluation function setting unit for calculating an evaluation function for integrating the smoothing values ​​while suppressing the increase in the error covariance matrix and guaranteeing an upper limit of the error covariance matrix due to errors in the correlation coefficient; and an optimal weight calculation unit for calculating weights using the evaluation function transmitted from the evaluation function setting unit.

7. The signal processing apparatus according to claim 6, characterized in that the optimal weight calculation unit calculates weights by applying the Newton method using the evaluation function.

8. The signal processing apparatus according to any one of claims 2 to 5, wherein the weight setting unit comprises: a memory circuit for storing observation data transmitted from the radar and the smoothing values ​​transmitted from the radar tracking filter; a correlation coefficient calculation unit for determining the correlation coefficient between the smoothing values ​​output from the radar tracking filters corresponding to different radars; an error covariance matrix calculation unit for calculating the error covariance matrix based on the correlation coefficient transmitted from the correlation coefficient calculation unit; a search-type evaluation function setting unit for determining an evaluation function for integrating the smoothing values ​​while suppressing the increase in the error covariance matrix by search; and a full-search type optimal weight calculation unit for calculating weights by performing a full search using the evaluation function transmitted from the search-type evaluation function setting unit.

9. The signal processing apparatus according to any one of claims 2 to 5, wherein the weight setting unit comprises: a memory circuit for storing observation data transmitted from the radar and the smoothing values ​​transmitted from the radar tracking filter; a correlation coefficient calculation unit for determining the correlation coefficient between the smoothing values ​​output from the radar tracking filters corresponding to different radars; an error covariance matrix calculation unit for calculating the error covariance matrix based on the correlation coefficient transmitted from the correlation coefficient calculation unit; a search-type evaluation function setting unit for determining an evaluation function for integrating the smoothing values ​​while suppressing the increase in the error covariance matrix through a search; and a range-limited search-type optimal weight calculation unit for calculating weights by performing a range-limited search using the evaluation function transmitted from the search-type evaluation function setting unit.

10. The signal processing device according to claim 9, characterized in that the range-limited search type optimal weight calculation unit searches for the weight for the current sampling only in the vicinity of the weight used in the previous sampling.

11. The signal processing apparatus according to any one of claims 6 to 10, characterized in that the correlation coefficient calculation unit calculates the correlation coefficient using observation data over several past samples.

12. The signal processing apparatus according to claim 6 or 7, characterized in that the optimal weight calculation unit sets any weight with a negative sign to 0 when calculating the weights, and further adjusts so that the sum of the weights becomes 1.

13. The signal processing apparatus according to any one of claims 2 to 5, wherein the weight setting unit comprises: a memory circuit for storing observation data transmitted from the radar and the smoothing values ​​transmitted from the radar tracking filter; a correlation coefficient calculation unit for determining the correlation coefficient between the smoothing values ​​output from the radar tracking filters corresponding to different radars; an error covariance matrix calculation unit for calculating the error covariance matrix based on the correlation coefficient transmitted from the correlation coefficient calculation unit; an evaluation function setting unit for calculating an evaluation function for integrating the smoothing values ​​while suppressing the increase in the error covariance matrix while guaranteeing an upper limit of the error covariance matrix due to errors in the correlation coefficient; and a sensor-selection type optimal weight calculation unit for calculating weights using only the remaining sensors, excluding sensors corresponding to weights with a negative sign, when calculating weights using the evaluation function transmitted from the evaluation function setting unit.

14. A signal processing method by a signal processing device, comprising: a step of a radar tracking filter corresponding to each of a plurality of radars calculating a smoothed value indicating the smoothed position of a target using observation data acquired by each radar, including the direction of existence of the target and the distance to the target, and outputting the calculated smoothed value and an error covariance matrix of the smoothed value; a step of a weight setting unit setting weights for integrating the smoothed values ​​while suppressing the increase of the output error covariance matrix; and a step of a data integration unit integrating the smoothed values ​​using the set weights and estimating the position of the target based on the obtained estimated value.

15. A program to cause a computer to function as a tracking filter for multiple radars, each corresponding to a multiple radar, which calculates a smoothed value indicating the smoothed position of a target using observation data acquired by each radar, including the direction of the target's existence and the distance to the target, and outputs the calculated smoothed value and the error covariance matrix of the smoothed value; a weight setting unit which sets weights for integrating the smoothed values ​​while suppressing the increase of the output error covariance matrix; and a data integration unit which integrates the smoothed values ​​using the set weights and estimates the position of the target based on the obtained estimate.

Images